Data Mining Techniques in Predictive Medicine: An Application in hemodynamic prediction for abdominal aortic aneurysm disease

Doaa Sami Khafaga^*1, Abdelhameed Ibrahim², S. K. Towfek³, Nima Khodadadi⁴

¹ Department of Computer Sciences, College of Computer and Information Sciences, Princess Nourah bint Abdulrahman University, P.O. Box 84428,

 Riyadh 11671, Saudi Arabia

² Computer Engineering and Control Systems Department, Faculty of Engineering, Mansoura University, Mansoura 35516, Egypt

³ Computer Science and Intelligent Systems Research Center, Blacksburg 24060, Virginia, USA

⁴Department of Civil and Architectural Engineering, University of Miami, Coral Gables, FL, USA

Emails: dskhafga@pnu.edu.sa; afai79@mans.edu.eg; sktowfek@jcsis.org; nima.khodadadi@miami.edu

Abstract

Due to its potential to enhance patient outcomes and ease individualized therapy, predictive medicine has received considerable interest in recent years. In this article we examine the use of data mining in predictive medicine, with a particular emphasis on hemodynamic prediction for abdominal aortic aneurysm (AAA) disease. In AAA, the abdominal aortic wall becomes weakened and may rupture, putting the patient's life in danger. Clinical decision making and treatment planning for AAA rely heavily on accurate hemodynamic prediction. For developing these predictive models for hemodynamic assessment, we use the well-known data mining techniques of Random Forest (RF) and AdaBoost. To capture complicated interactions, the RF approach employs a collection of decision trees, while AdaBoost iteratively improves the model by giving more weight to examples that were incorrectly classified. The experimental evidence shows that these methods are effective in providing reliable estimates of the hemodynamics of AAA. This research adds to the expanding field of predictive medicine by providing new understanding of the potential of data mining methods to improve the quality of care for patients with AAA illness.

Keywords: Predictive medicine; Data mining; Hemodynamic prediction; Abdominal aortic aneurysm (AAA); Random Forest; AdaBoost.